Quote:
Originally Posted by JAMF
There are other ways to achieve craters. A basic double pyramid with 6 sides only costs 16 triangles. Inverted, that looks like a hole. Graphic cards of today have a trick called tessellation. If the crater object was marked as an object that would receive tessellation, the card would increase that 16 triangles to 48, fo example. The crater now looks much smoother and the circular hole will have 12 sides.
Now add another trick, normal mapping ( Dot3 bump mapping). Simplified, it's a texture, which tells the card to add extra height/thickness to a point on a model. Say white is very high/thick and black is nothing/zero. Lay a black&white noise texture over the crater and you get the inside surface to look like it's just exploded and it's covered with clumps of dirt and sand.
These effects should have little effect on the frame rate, as they can be distance-indexed, so they start to show only when you get closer. Similar to the LoD models aircraft have.
Bump mapping is an optical illusion, normal mapping really adds surface detail.
Left is bump-, right is normal-mapped. Notice the visual edge of the spheres:
 |
My understanding is that normal mapping is simply bump mapping with higher precision (explicit vs. implicit surface normals) but you still don't change the polygon shape of the object, and the shading is confined to the polygon area (no additional detail visible at the edges). That's where tesselation would come into play, I think that's what would be required to get the result on the image to the right. Could be wrong though; I've done some graphics programming but it's been a while
Using displacement mapping could work nicely with craters, as this adds more depth than bump or normal maps and there'd still be one flat texture per crater like in Il2. Using actual 3D objects is still difficult because they have to go through a separate rendering pass or they'll interfere with the land surface (the crater surface has to be under the ground level). And making the landscape itself deformable with the resolution to render craters would be even more difficult. But who knows, Oleg may yet surprise us. I guess he's already chosen a technique so it's a bit late anyway to give suggestions
